Past events

Our brain relies on internal representations to support perception, action, and decision-making. Internal representations are usually rich, multidimensional, and cannot be directly observed. How can these internal representations be characterized? How are they affected by experience? My work develops adaptive behavioral paradigms that integrate human decisions into computer algorithms via human-in-the-loop experiments. I combine these paradigms with a data-intensive expansion of the scale and scope of behavioral research by means of massive online experiments and cross-cultural comparative research. This talk presents “adaptive sampling,” a type of experimental paradigm inspired by Monte Carlo Markov Chain techniques. Each successive stimulus depends on a subject's response to the previous stimulus. This process allows us to sample from the complex and high-dimensional joint distribution associated with internal representations and obtain high-resolution maps of perceptual spaces. After introducing these methods and describing their implementation via large-scale online experiments and field experiments around the world, I demonstrate how they can be applied to fundamental questions in the understanding of the human mind. Specifically, I examine how biology and culture influence internal representations and how semantics influence perception.

Humans and animals tend to behave differently when learning from rewarding or aversive feedback, and the amygdala is hypothesized to play a role in these differences. Here, we studied neural mechanisms of learning and decision making in reward and punishment, namely post-stimulus rehearsal, balancing of exploration and exploitation and generalization. To study post-stimulus rehearsal in amygdala neurons, we investigated spike-sequences across simultaneously recorded neurons of non-human primates, while they learned to discriminate between aversive and pleasant tone-odor associations. We showed that valence specific sequences across amygdala neurons rehearsed the coding of the recent association, so they can serve as a coding mechanism that enhances memory formation by rehearsal of the recent association. Next, to examine neural coding of exploration under rewards and punishments, we recorded single neurons while human subjects were engaged in a probabilistic decision-making task with gain and loss conditions, and found more exploration when subjects tried to minimize their losses. We found two mechanisms of explorational choices: one is executed through firing rate of single neurons in the temporal cortex and amygdala and is shared across valence, and the other is executed by an increase in noise in amygdala neurons, and is specific to the loss condition. Finally, we found that over-generalization around a loss-conditioned tone was accompanied by a similar over response of amygdala neurons. Together, this work expands the knowledge of neural mechanisms that enhance learning and improves decision making, specifically in complex environments that include opportunities for rewards and risks for punishments. Zoom link: https://weizmann.zoom.us/j/96622589021?pwd=Tkh1RWk0OFhaVFE0SW9KeU84Q1cvZz09 Meeting ID: 966 2258 9021 Password: 022196

Populations of hippocampal neurons have been hypothesized to operate collectively to support the stable maintenance of long-term memories. To test this hypothesis, we performed large-scale calcium imaging in the hippocampus of freely behaving mice that repeatedly explored the same environments over weeks. Surprisingly, we discovered that across separate visits to the same familiar environment, hippocampal neurons can collectively switch between multiple distinct spatial representations, without any apparent changes in sensory input or animal’s behavior. The distinct representations were spatially informative and stable over weeks, and switching between them required a complete disconnection of the animal from the environment, demonstrating the coexistence of distinct stable attractors in the hippocampal network. In the second part of the talk, I will present a comparison of the coding properties between hippocampal subfields CA1 and CA3 in novel environments. Place cells in CA3 had more precise and stable spatial tuning than place cells in CA1. Moreover, we showed that in CA3 the tuning of place cells exhibited a higher statistical dependence with their peers compared to in CA1, uncovering an organization of CA3 into cell assemblies. Interestingly, cells with stronger tuning peer-dependence had higher stability but not higher precision, suggesting that distinct mechanisms control these two aspects of the neural code. Overall, our results demonstrate that multiple attractor states can stably coexist in the hippocampus and suggest that a cell-assembly organization in hippocampal CA3 underlies the long-term maintenance of stable spatial codes. Link:https://weizmann.zoom.us/j/97167587409?pwd=TDFFYWI0ZmF5YXk0TW5oN1ZKSStndz09 Meeting ID: 971 6758 7409 Password: 227875

How do fleeting molecules and dynamic neural codes enable the conversion of transient stimuli into lasting internal representations? And are there unique strategies to achieve memory on different time scales. Our lab addresses these questions by bridging functional genomics with systems neuroscience to provide cross-disciplinary insights. On one hand, we perform genetic mapping in outbred mice for unbiased discovery of genes, cell types, and circuits relevant for memory across different time scales. In parallel, we develop and apply methodologies to record and manipulate high resolution neural activity from these relevant circuits in the behaving animal. In today’s talk, I will discuss how these approaches have led to new insights into the genetic contributions and long-range circuit dynamics that facilitate both short- and long- term memory.  Zoom Link: https://weizmann.zoom.us/j/95406893197?pwd=REt5L1g3SmprMUhrK3dpUDJVeHlrZz09 Meeting ID: 954 0689 3197 Password: 750421

Our brain relies on internal representations to support perception, action, and decision-making. Internal representations are usually rich, multidimensional, and cannot be directly observed. How can these internal representations be characterized? How are they affected by experience? My work develops adaptive behavioral paradigms that integrate human decisions into computer algorithms via human-in-the-loop experiments. I combine these paradigms with a data-intensive expansion of the scale and scope of behavioral research by means of massive online experiments and cross-cultural comparative research. This talk presents “adaptive sampling,” a type of experimental paradigm inspired by Monte Carlo Markov Chain techniques. Each successive stimulus depends on a subject's response to the previous stimulus. This process allows us to sample from the complex and high-dimensional joint distribution associated with internal representations and obtain high-resolution maps of perceptual spaces. After introducing these methods and describing their implementation via large-scale online experiments and field experiments around the world, I demonstrate how they can be applied to fundamental questions in the understanding of the human mind. Specifically, I examine how biology and culture influence internal representations and how semantics influence perception.

Humans and animals tend to behave differently when learning from rewarding or aversive feedback, and the amygdala is hypothesized to play a role in these differences. Here, we studied neural mechanisms of learning and decision making in reward and punishment, namely post-stimulus rehearsal, balancing of exploration and exploitation and generalization. To study post-stimulus rehearsal in amygdala neurons, we investigated spike-sequences across simultaneously recorded neurons of non-human primates, while they learned to discriminate between aversive and pleasant tone-odor associations. We showed that valence specific sequences across amygdala neurons rehearsed the coding of the recent association, so they can serve as a coding mechanism that enhances memory formation by rehearsal of the recent association. Next, to examine neural coding of exploration under rewards and punishments, we recorded single neurons while human subjects were engaged in a probabilistic decision-making task with gain and loss conditions, and found more exploration when subjects tried to minimize their losses. We found two mechanisms of explorational choices: one is executed through firing rate of single neurons in the temporal cortex and amygdala and is shared across valence, and the other is executed by an increase in noise in amygdala neurons, and is specific to the loss condition. Finally, we found that over-generalization around a loss-conditioned tone was accompanied by a similar over response of amygdala neurons. Together, this work expands the knowledge of neural mechanisms that enhance learning and improves decision making, specifically in complex environments that include opportunities for rewards and risks for punishments. Zoom link: https://weizmann.zoom.us/j/96622589021?pwd=Tkh1RWk0OFhaVFE0SW9KeU84Q1cvZz09 Meeting ID: 966 2258 9021 Password: 022196

Populations of hippocampal neurons have been hypothesized to operate collectively to support the stable maintenance of long-term memories. To test this hypothesis, we performed large-scale calcium imaging in the hippocampus of freely behaving mice that repeatedly explored the same environments over weeks. Surprisingly, we discovered that across separate visits to the same familiar environment, hippocampal neurons can collectively switch between multiple distinct spatial representations, without any apparent changes in sensory input or animal’s behavior. The distinct representations were spatially informative and stable over weeks, and switching between them required a complete disconnection of the animal from the environment, demonstrating the coexistence of distinct stable attractors in the hippocampal network. In the second part of the talk, I will present a comparison of the coding properties between hippocampal subfields CA1 and CA3 in novel environments. Place cells in CA3 had more precise and stable spatial tuning than place cells in CA1. Moreover, we showed that in CA3 the tuning of place cells exhibited a higher statistical dependence with their peers compared to in CA1, uncovering an organization of CA3 into cell assemblies. Interestingly, cells with stronger tuning peer-dependence had higher stability but not higher precision, suggesting that distinct mechanisms control these two aspects of the neural code. Overall, our results demonstrate that multiple attractor states can stably coexist in the hippocampus and suggest that a cell-assembly organization in hippocampal CA3 underlies the long-term maintenance of stable spatial codes. Link:https://weizmann.zoom.us/j/97167587409?pwd=TDFFYWI0ZmF5YXk0TW5oN1ZKSStndz09 Meeting ID: 971 6758 7409 Password: 227875

How do fleeting molecules and dynamic neural codes enable the conversion of transient stimuli into lasting internal representations? And are there unique strategies to achieve memory on different time scales. Our lab addresses these questions by bridging functional genomics with systems neuroscience to provide cross-disciplinary insights. On one hand, we perform genetic mapping in outbred mice for unbiased discovery of genes, cell types, and circuits relevant for memory across different time scales. In parallel, we develop and apply methodologies to record and manipulate high resolution neural activity from these relevant circuits in the behaving animal. In today’s talk, I will discuss how these approaches have led to new insights into the genetic contributions and long-range circuit dynamics that facilitate both short- and long- term memory.  Zoom Link: https://weizmann.zoom.us/j/95406893197?pwd=REt5L1g3SmprMUhrK3dpUDJVeHlrZz09 Meeting ID: 954 0689 3197 Password: 750421